the long and short of the canada-u.s. free trade … › ~dtrefler › papers ›...

The Long and Short of the Canada-US Free Trade Agreement

By DANIEL TREFLER

The Canada-US Free Trade Agreement provides a unique window onto the effectsof a reciprocal trade agreement on an industrialized economy (Canada) Forindustries that experienced the deepest Canadian tariff cuts the contraction oflow-productivity plants reduced employment by 12 percent while raising industry-level labor productivity by 15 percent For industries that experienced the largestUS tariff cuts plant-level labor productivity soared by 14 percent These resultshighlight the conflict between those who bore the short-run adjustment costs(displaced workers and struggling plants) and those who are garnering the long-rungains (consumers and efficient plants) (JEL F13 F14 F15 F16 D24)

The central tenet of international economicsis that free trade is welfare improving We ex-press our conviction about free trade in ourtextbooks and we sell it to our politicians Yetthe fact of the matter is that we have one heckof a time explaining these benefits to the largerpublic a public gripped by Free Trade Fatigue

Why is the message of professional econo-mists not more persuasive To my mind thereare two reasons First in examining trade liber-

alization we treat short-run transition costs andlong-run efficiency gains as entirely separateareas of inquiry On the one hand are those whostudy the long-run productivity benefits of freetrade policies eg James R Tybout et al(1991) James Levinsohn (1993) Ann E Har-rison (1994) Tybout and M Daniel Westbrook(1995) Pravin Krishna and Devashish Mitra(1998) Keith Head and John Ries (1999a b)and Nina Pavcnik (2002) On the other hand arethose who study the impacts of freer trade onshort-run worker displacement and earningseg Noel Gaston and Trefler (1994 1995) AnaRevenga (1997) Levinsohn (1999) EugeneBeaulieu (2000) and Krishna et al (2001)Only Janet Currie and Harrisonrsquos (1997) studyof Morocco examines both labor market out-comes and productivity In assessing free tradepolicies there is clearly a bias introduced whenlooking only at the long-run benefits or only atthe short-run costs Nowhere is this more ap-parent than for the Canadian experience withthe Canada-US Free Trade Agreement (FTA)and its extension to Mexico The FTA triggeredongoing and heated debates about freer tradeThis heat was generated by the conflict betweenthose who bore the short-run adjustment costs(displaced workers and stakeholders of closedplants) and those who garnered the long-runefficiency gains (stakeholders of competitiveplants and users of final and intermediategoods)

There is another reason why the free trade mes-sage is not more persuasive While case-study

Joseph L Rotman School of Management and Depart-ment of Economics University of Toronto 105 St GeorgeStreet Toronto Ontario Canada M5S 3E6 Canadian In-stitute for Advanced Research (CIAR) and the NationalBureau of Economic Research (NBER) (e-mail dtreflerrotmanutorontoca) I have benefited from the research as-sistance of Yijun Jiang Huiwen Lai Runjuan Liu andSusan Zhu Michael Baker Erwin Diewert Peter DunganGerry Helleiner Pravin Krishna Aloysius Siow and mem-bers of the CIAR provided key comments that dramaticallyimproved the paper Much more than is usually the casethis paper was dramatically overhauled in response to thecreative input from two anonymous referees Many mem-bers of Statistics Canada provided advice on data issuesincluding Richard Barnabe (Standards Division) JocelyneElibani (International Trade Division) Richard Landry (In-vestment and Capital Stock Division) Jean-Pierre Maynard(Productivity Division) Bruno Pepin (Industry Division)and Bob Traversy (Industry Division) I am especially grate-ful to John Baldwin (Director Micro-Economic Studies andAnalysis Division) for making available the plant-leveldata Alla Lileeva was kind enough to analyze the plant-level data in Ottawa Research support from the SocialSciences and Humanities Research Council of Canada(SSHRC) is gratefully acknowledged Views expressed inthis paper do not necessarily reflect those of Statistics Can-ada or SSHRC

870

evidence abounds about efficiency gains fromliberalization (eg Anne O Krueger 1997)solid econometric evidence for industrializedcountries remains scarce When I teach my stu-dents about the effects of free trade on produc-tivity I turn to high-quality studies for Chile(Tybout et al 1991 Pavcnik 2002) Turkey(Levinsohn 1993) Cote drsquoIvoire (Harrison1994) Mexico (Tybout and Westbrook 1995)and India (Krishna and Mitra 1998) amongothers I find these studies compelling but Iwonder whether they can be expected to per-suade policy makers and voters in industrializedcountries such as Canada and the United StatesWhat is needed is at least some research focus-ing on industrialized countries

The Canada-US Free Trade Agreement of-fers several advantages for assessing the short-run costs and long-run benefits of tradeliberalization in an industrialized country Firstthe FTA policy experiment is clearly defined Indeveloping countries trade liberalization is typ-ically part of a larger package of market re-forms making it difficult to isolate the role oftrade policy Further the market reforms them-selves are often initiated in response to majormacroeconomic disturbances Macroeconomicshocks market reforms and trade liberalizationare confounded Indeed Gerald K Helleiner(1994 p 28) uses this fact to argue that ldquoEm-pirical research on the relationship between to-tal factor productivity (TFP) growth and thetrade regime has been inconclusiverdquo His viewis widely shared eg Harrison and Gordon HHanson (1999) and Francisco Rodriguez andDani Rodrik (2001) In contrast the FTA wasnot implemented as part of a larger package ofreforms or as a response to a macroeconomiccrisis Second as Harrison and Revenga (1995abstract) note ldquoTrade policy is almost nevermeasured using the most obvious indicatorsmdashsuch as tariffsrdquo Tybout (2000) echoes this crit-icism My study of the FTA is particularlycareful about constructing pure policy-mandatedtariff measures

Third the FTA is not just about import-liberalizing policies It is a reciprocal agreementthat includes export-liberalizing policies aswell It should therefore be expected to induce apronounced general-equilibrium relocation ofresources out of import-competing sectors and

into export-oriented sectors I will examinethese FTA effects on a large number of Cana-dian plant and industry outcomes At the plantand industry levels the outcomes include em-ployment and earnings of both production andnonproduction workers skill upgrading earn-ings inequality hours of work plant size andlabor productivity

Fourth the FTA is a preferential trading ar-rangement Such arrangements need not bewelfare-improving I will examine the two con-ditions usually put forward as sufficientmdashatleast informallymdashfor welfare gains These arethat trade creation must dominate trade diver-sion and that import prices must not rise(Arvind Panagariya 2000 Krishna 2003) Bothconditions are satisfied

The backdrop of the FTAmdashan industrializedcountry a clean policy experiment the directpolicy lever of tariffs general-equilibrium rec-iprocity effects and the long list of outcomesincluding employment productivity and pricesmdashwill be my basis for a rigorous and detailedexamination of the short-run costs and long-runbenefits of trade liberalization

The FTA has been the subject of severalstudies since its implementation on January 11989 Gaston and Trefler (1997) found that theFTA had no effect on earnings and only a mod-est effect on employment Beaulieu (2000)found that the employment effect was primarilydriven by modest nonproduction worker em-ployment losses Kimberly A Clausing (2001)found evidence that the FTA raised US im-ports from Canada (trade creation) but did notdivert US imports away from other US trad-ing partners John Romalis (2004) found bothtrade creation and diversion The most intrigu-ing FTA study is by Head and Ries (1999b)They found that the FTA had little net effecton industry-level average output per plant(which they take as a proxy for scale) and apuzzling effect on Canadian plant exit (exitwas induced by falling Canadian tariffs andby falling US tariffs) Unfortunately noneof these papers uses plant-level data FurtherI will argue below that at least some of thesepapers (including my own) suffer specifica-tion issues that substantively mar the infer-ences drawn about the effects of the Canada-US Free Trade Agreement

871VOL 94 NO 4 TREFLER CANADA-US FREE TRADE

I The FTA Tariff Cuts Too Small to Matter

This paper deals with the impact of FTA-mandated tariff cuts The top panel of Figure 1plots Canadarsquos average manufacturing tariffagainst the United States (solid line) and Can-adarsquos average manufacturing tariff against therest of the world (dashed line) The bottompanel plots the corresponding US tariffsagainst Canada (solid line) and the rest of theworld (dashed line) In 1988 the average Cana-dian tariff rate against the United States was 81percent The corresponding effective tariff ratewas 16 percent1 Perhaps most importantly tar-

iffs in excess of 10 percent sheltered one in fourCanadian industries Given that these industrieswere almost all characterized by low wageslow capital-labor ratios and low profit marginsthe 1988 tariff wall was indeed high Similarcomments apply to the US tariff against Can-ada albeit with less force since the average1988 US tariff was 4 percent

That one in four Canadian industries hadtariffs in excess of 10 percent depends cruciallyon the level of aggregation I am working with4-digit Canadian SIC data (213 industries) Ifone aggregates up even to 3-digit data (105industries) almost no industries had 1988 tariffsin excess of 10 percent This is important be-cause studies of trade liberalization typically donot work with such disaggregated tariff dataFor example papers by Tybout et al (1991)Levinsohn (1993) Harrison (1994) Tybout andWestbrook (1995) Gaston and Trefler (1997)Krishna and Mitra (1998) and Beaulieu (2000)are never at a finer level of aggregation than3-digit ISIC with its 28 manufacturing sectors

The core feature of the FTA is that it reducedtariffs between Canada and the United Stateswithout reducing tariffs against the rest of theworld Graphically the FTA placed a gap be-tween the dashed and solid lines of Figure 1Letting i index industries and t index years mymeasures of the FTA policy levers will be

itCA The FTA-mandated Canadian tariff

concessions granted to the UnitedStates In terms of the top panel ofFigure 1 this is the solid line minusthe dashed line

itUS The FTA-mandated US tariff

concessions granted to Canada Interms of the bottom panel of Figure1 this is the solid line minus thedashed line

itCA and it

US capture the core textual aspects ofthe FTA2

1 Both the nominal and effective tariff rates were aggre-gated up from the 4-digit SIC level using Canadian produc-tion weights The standard formula used to calculate theeffective rate of protection appears in Trefler (2001 p 39)Details about construction of the tariff series appear inAppendix A

2 Given that tariffs are positively correlated with effec-tive tariffs and nontariff barriers to trade (NTBs) thecoefficients on it

CA and itUS will capture the effects of

FTA-mandated reductions in tariffs effective tariffs andnontariff barriers This is exactly what I want When

FIGURE 1 CANADIAN AND US BILATERAL TARIFFS IN

MANUFACTURING

(In Percents)

872 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004

II Econometric Strategy

In this section I lay out econometric strate-gies for analyzing the plant- and industry-leveldata I begin with the latter Let i index indus-tries let t index years and let Yit be a Canadianoutcome of interest such as employment or pro-ductivity The FTA mandates that tariffs bereduced once a year on January 1 starting in1989 I have data for the FTA period 1989ndash1996 In what follows I will define the pre-FTAperiod as the years 1980ndash1986 As will beshown in detail this choice is useful for dealingwith business fluctuations Let yis be the aver-age annual log change in Yit over period s wheres 1 indexes the FTA period and s 0 indexesthe pre-FTA period That is

(1) yi1 ln Yi1996 ln Yi19881996 1988

and

yi0 ln Yi1986 ln Yi19801986 1980

The FTA period changes use 1988 data becauseI am interested in comparing the FTA-periodoutcome Yi1996 with its baseline level ie withits level before the first round of tariff reduc-tions on January 1 19893 For k CA and k US define

(2) i1k i1996

k i1988k 1996 1988

i1CA measures the change in the FTA-

mandated tariff concessions extended by Can-ada to the United States Likewise i1

US

measures the change in the FTA-mandated tariffconcessions extended by the United States toCanada

What of pre-FTA period tariff concessionswhich I denote by i0

k Except for the 1965Canada-US Auto Pact all tariff rates wereextended on a Most Favored Nation (MFN)basis prior to 1988 Thus define i0

k (i1986

k i1980k )(1986 1980) when industry

i is in the automotive sector and i0k 0

otherwise As will be shown setting i0k 0

for all i or omitting the automotive sector en-tirely from the analysis makes no difference tothe results Additional details about i1

k including a list of industries with large abso-lute values of i1

CA and i1US appear in Appen-

dix AI am interested in a regression model that

explains the impact of the FTA-mandated tariffconcessions on a variety of industry outcomes

(3) yis s CAisCA

USisUS is s 0 1

where s is a period fixed effect There is anobvious problem with estimating equation (3) Ihave no deeply satisfying way of controlling forthe lack of randomization in the tariff conces-sions I must thus take particular care to controlboth for the endogeneity of tariffs and forsources of industry-level heterogeneity thatmight contaminate the estimates of CA andUS I turn to this task now

A The Secular Growth Control

For political economy reasons one might ex-pect declining industries to have high tariffs andhence deep FTA tariff concessions eg Trefler(1993) To prevent mistakenly attributing secu-lar growth trends to the FTA tariff concessionsI introduce a growth fixed effect i into equa-tion (3)

(4) yis i s CAisCA

USisUS is s 0 1

As a result CA and US only pick up FTA

analyzing tariff concessions I am actually capturing abroader set of FTA trade-liberalizing policies

3 Since this may cause some confusion consider byanalogy a cholesterol-reducing drug trial in which the drugis given once a year on January 1 (starting in 1989) and thepatientrsquos cholesterol level Yit is measured once a year onDecember 31 (starting in 1988) To measure the long-termeffects of the drug one looks at Yi1996 Yi1988 rather thanYi1996 Yi1989 because Yi1988 describes the patient cho-lesterol baseline without drugs The same logic holds for theldquodrug of free traderdquo The FTA mandates that tariffs bereduced once a year on January 1 (starting in 1989) and theplants are surveyed once a year as closely as possible toDecember 31 Therefore the appropriate baseline is Yi1988


impacts on industry growth that are departuresfrom industry trend growth

B Industry-Specific Shocks

A number of Canadian industries experi-enced reversals of fortune in the sense thatemployment growth in the pre-FTA and FTAperiods had opposite signs For these industriessimilar reversals also appeared in their UScounterparts This is indicative of industry-specific demand and supply shocks If thesereversals of fortune are a characteristic of highlyprotected industries the reversals might con-taminate the estimates of CA and US Control-ling for reversals of fortune begins with theobservation that many industry-specific shocksthat appeared in Canada also appeared in Can-adarsquos major trading partners For examplehigher oil prices affected the petroleum industryin Canada and all its major trading partners Ihave industry-level data for Canadarsquos three larg-est trading partners the United States Japanand the United Kingdom I use these data tocontrol for industry-specific shocks

More formally let y isj be data on yis for

economy j eg if yis is Canadian employmentgrowth then y is

j is country jrsquos employmentgrowth I control for industry-specific shocks byincluding y is

j in equation (4) Note that y isj

may be endogenous especially for j US so Iwill employ instrumental variables (IV) tech-niques Finally for expositional ease I will referto y is

j as the ldquoUS controlrdquo and simply writeyis

US

C The Business Conditions Control

A key issue for examining the FTA is thetreatment of the early 1990rsquos recession Figure 2plots GDP in year t for Canadian manufacturing(gdpt) The data are in logs relative to a 1980base ie ln(gdpt gdp1980) The FTA periodrecession stands out This is a problem if theindustries that experienced the deepest tariffconcessions share a common sensitivity tochanges in business conditions General busi-ness conditions can be introduced into equation(4) by including a regressor bis that captureshow movements in GDP and the real exchangerate affect industry i I will explain how bis is

constructed shortly Introducing bis and yisUS

into equation (4) yields

(5) yis i s CAisCA USis

US

yisUS bis is s 0 1

D Estimation

Differencing (5) across periods yields mydifference-of-differences baseline specification

(6) yi1 yi0 CAi1CA i0

CA

USi1US i0

US

yi1US yi0

US

bi1 bi0 i

where 1 0 This specification controlsfor secular industry trends (by differencing outthe i) industry-specific demand and supplyshocks (the yis

US) and industry-specific busi-ness condition effects (the bis) Clearly Iwill have to use an IV estimator to deal withthe endogeneity of the tariff concessions andyi1

US yi0US

It is important to note that the use of longdouble-differencing means that I need not worryabout dynamic panel estimation problems(Manuel Arellano and Bo Honore 2001) Thisis important because every single previous FTAstudy has used annual data without any correc-tion for autocorrelation eg Gaston and Trefler

FIGURE 2 REAL CANADIAN MANUFACTURING GDP

Note gdp at fa4ctor cost 1992 dollars


(1997) Head and Ries (1999a b) Beaulieu(2000) and Clausing (2001) Yet the fact is thatemployment and output display strong autocor-relation at lags of up to three years For exam-ple Canadian employment displays significantthree-year autocorrelation in 31 percent of allindustries and one-year autocorrelation in anoverwhelming 77 percent of all industriesThus the estimators used in all previous studiesof the FTA (including my own) are inconsistentand yield standard errors that are too small

E Plant-Level Data

Letting k index plants my baseline plant-level specification is

(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik

where yiks is the change in the outcome ofinterest for plant k in industry i in period s andxik1980 is a vector of plant characteristics thatincludes the log of 1980 employment the log of1980 earnings per worker the log of 1980 laborproductivity and the log of plant age Since theplant data only go back to 1973 I also includea dummy for whether the plant was older thanseven years of age in 1980 There are 3801plants in the sample4

There are two selection issues that requireattention First equation (7) only makes use ofplants that were in existence in 1980 19861988 and 1996 Obviously these ldquocontinuingrdquoplants are not representative of all plants Un-fortunately I have not been able to make evensimple corrections for entry and exit because thedatabase available to me cannot be used in anysimple way to track entry and exit (Unlike theUS longitudinal plant database the Canadian

database has not attracted as many resources fordata ldquocleaningrdquo) Second I will be working withwhat are known as ldquolong-formrdquo plants that isplants that fill out a detailed survey In 1988long-form plants were 22 times larger thanldquoshort-formrdquo plants Thus my plant-level re-sults must be understood as dealing with largerplants This said Appendix E provides someevidence that my results apply to small plants aswell5

III The Data

Canadian data are from the Canadian AnnualSurvey of Manufactures (ASM) the CanadianLabour Force Survey as well as StatisticsCanadarsquos International Trade Division Input-Output Division Prices Division and StandardsDivision (for commodity and industry concor-dances) Almost all the data used involved spe-cial tabulations by Statistics Canada Most ofthe US data through 1994 are from the NBERManufacturing Productivity Database (Eric JBartelsman and Wayne Gray 1996) and fromRobert C Feenstra (1996) I updated thesesources to 1996 As discussed in Trefler (2001p 11) I have been especially careful to build aCanada-US converter that steps down fromover 1000 US products to 213 Canadianindustries

IV Empirical Results Employment

Table 1 reports estimates of equations (6) and(7) for the case where the dependent variable isemployment growth The table includes a largenumber of specifications in order to show thatthe estimates of CA and US are not particularlysensitive to the choice of specification Row 1 ismy industry-level baseline specification It usesordinary least squares (OLS) and includes allfour regressors I will explain coefficient

4 I am indebted to Alla Lileeva for running these regres-sions and for sharing her experience as to which plant-levelcontrols to use Without her the plant-level analysis wouldnot have been possible

5 One final thought on the estimating equation Thispaper is unabashedly a reduced-form exercise that allowsthe inferences to be driven more by the data than by a highlystructured model This has obvious advantages but it alsohas a cost A more structured approach as in Head and Ries(2001) or Huiwen Lai and Trefler (2002) muzzles the databut allows for a clearer interpretation of the coefficients andfor a richer treatment of general-equilibrium feedbacks


magnitudes shortly but for now treat CA andUS as the log-point changes in employmentassociated with the FTA For example the Ca-nadian tariff concessions led to a 012 log-point change in employment (t 235)

The first specification issue handled by Table1 deals with the sensitivity of CA and US tothe way in which the business conditions vari-able bis is constructed In order to explain howbis is constructed define zt (ln gdpt ln rert)where rert is the real exchange rate and let 1 bethe annual difference operator so that 1zt zt zt1 and 1yit yit yit1 To constructbis I first regressed 1yit on (1zt 1ztJ)for some lag length J This is a time-seriesregression that was estimated separately for

each i The regression generates an industry-specific prediction 1 yit of the effect of currentand past business conditions on current annualemployment growth Second note from equa-tion (1) that yi1 can be written as yent1989

1996

1yit 8 This motivates the definition of bi1 asbi1 yent1989

1996 1 yit 8 bi1 is just an industry-specific prediction of the effect of business con-ditions on FTA-period employment growth Forthe pre-FTA period I use bi0 yent1981

1986

1 yit 6 Note that there is a different bis foreach outcome For example when yis is earn-ings growth then bis is the portion of industryi earnings growth driven by movements in GDPand the real exchange rate See Appendix C forfurther details

TABLE 1mdashDETAILED RESULTS FOR EMPLOYMENT

Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t

Industry level OLS1 gdp rer (2) 012 235 003 067 029 696 015 221 024 005 2662 gdp rer (0) 011 203 004 091 030 366 021 275 012 006 2583 gdp (2) 011 208 003 066 037 660 015 216 023 005 2414 mdash 014 240 002 052 020 258 007 006 2585 gdp rer (2) 013 248 002 039 028 674 029 300 024 005 1716 gdp rer (2) 014 275 003 080 030 712 023 006 3167 mdash 017 288 003 066 004 007 3158 gdp rer (2) 014 224 002 053 029 689 015 211 024 006 2659 gdp rer (2) 012 230 006 145 030 723 014 204 027 006 324

Plant level OLS10 gdp rer (2) 012 376 000 015 013 459 025 529 004 004 32611 gdp rer (2) 012 360 001 026 016 563 025 521 002 004 351Industry level IV12 gdp rer (2) 024 145 009 066 029 668 015 206 022 060065 004 12613 gdp rer (2) 024 143 004 029 031 637 016 050 020 067057 005 157Plant level IV14 gdp rer (2) 019 240 007 094 013 430 024 496 004 014099 004 25515 gdp rer (2) 019 244 007 092 013 417 016 095 003 010089 004 310

Notes The dependent variable is the log of employment The estimating equation is equation (6) for the industry-levelregressions and equation (7) for the plant-level regressions CA is scaled so that it gives the log-point impact of the Canadiantariff concessions on employment in the most impacted import-competing industries US is scaled so that it gives thelog-point impact of the US tariff concessions on employment in the most impacted export-oriented industries The ldquoTotalFTA impactrdquo column gives the joint impact of the tariff concessions on employment in all 213 industries The ldquoOverIdHausmanrdquo column reports p-values for the overidentification and Hausman tests Rejection of the instrument set or exogeneityare indicated by p-values less than 001 The number of observations is 213 for the industry-level regressions and 3801 forthe plant-level regressions In rows 4 and 7 the business conditions variable is omitted so that business conditions arecontrolled for implicitly by double-differencing yi1 yi0 In row 5 the US control is replaced by the Japan-UK controldiscussed in the text In row 8 the 2 ldquooutlierrdquo observations with the largest Canadian tariff cuts are omitted In row 9 all 9observations associated with the automotive sector are omitted In row 11 the plant controls are omitted In rows 12 and 14only the Canadian and US tariff variables are instrumented In rows 13 and 15 the two tariff variables and the US controlare instrumented


Row 1 of Table 1 uses my baseline specifi-cation of bis in which the lag length is J 2I chose J 2 because the industry-specific auto-correlation functions only vanish at longer lagsRow 2 of Table 1 which uses J 0 illustratesthat CA and US are not sensitive to the choiceof lag length Row 3 uses J 2 but drops thereal exchange rate (rert) from zt This does notdramatically alter the estimates either In fact asrow 4 shows the estimates rise only slightlywhen bi1 bi0 is omitted from the base-line specification This requires some expla-nation as it might be misinterpreted to meanthat business conditions are playing only aminor role

Returning to Figure 2 the 1980ndash1986 and1988ndash1996 periods are very similar in terms ofbusiness conditions Each began a year beforethe peak each entered a deep recession in thethird year and each ended in the midst of aprolonged expansion Further my decision toend the pre-FTA period in 1986 ensures that thetwo periods are similar as judged by GDPgrowth over the period and by the number ofyears into the expansion That is I have pur-posely chosen the pre-FTA period so that afterdouble-differencing my estimating equationshave a built-in implicit control for businessconditions This explains why omitting bi1 bi0 does not dramatically alter the results Alsonote that the results are similar with the pre-FTA period defined as 1980ndash1988 or the FTAperiod defined as 1988ndash1994 See AppendixTable A2

Finally bi1 bi0 is a generated regressorwhich means that some care is needed to ensurecorrect standard errors Fortunately it isstraightforward to show that my reported OLSstandard errors come from the same distributionas the asymptotically ldquotruerdquo (ie N-limiting)distribution This can be shown by verifyingthat condition (63) of Jeffrey M Wooldridge(2002 p 116) is satisfied Further specificationtests are discussed in Appendix C

Consider now the US control variable yi1US

yi0US Its coefficient is positive for almost all

results reported in this paper This is to beexpected if it is picking up demand and supplyshocks that are common to both US and Ca-nadian industries Row 5 replaces yi1

US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2

Comparison of row 5 with row 1 reveals that thismakes little difference to CA or US Row 6shows that the omission of the US control alsomakes little difference Clearly CA and US arenot sensitive to how the US control is modeledThis conclusion will continue to hold when I in-strument the US control in row 136

Row 7 shows that omission of both the UScontrol and the business conditions control hasno effect on US but does lower CA from012 to 017 I conclude from rows 1ndash7 thatmy row 1 baseline estimates are not sensitive tothe exact treatment of industry-specific shocks(the US control) or the business conditionscontrol provided that at least one of them isincluded in the specification This conclusionholds true for all the statistically significantestimates reported in this paper

Rows 8 and 9 examine the role of particularobservations As Appendix Table A1 shows theBrewery and Shipbuilding industries have un-usually large Canadian tariff concessions andare thus potentially influential observations Inrow 8 I delete these observations This slightlyraises CA In row 9 I delete the nine industriesin the automotive sector This raises US butnot significantly

Row 10 is my baseline plant-level specifica-tion It includes the plant-level controls ieplant age and the 1980 values of the log ofemployment the log of earnings and the log oflabor productivity Notice that the plant-levelestimates of CA and US are almost identical tothe industry-level estimates of row 1 Thissuggests that at least for employment the

6 Throughout this paper I will use US data rather thanJapanndashUK data The disadvantage of using yis

US is that theCanadian tariff concessions likely raised US employmentat the expense of Canadian employment However if thiswere an important feature of the data then I would expectthe correlation between yi1

US and yi1 to be negative (in factit is a strongly positive 050) and the coefficient on (yi1

US yi0

US) to be negative (in fact it also is strongly positive)The disadvantage of (yis

Japan yisUK) 2 is that these

data are only available at the 3-digit ISIC level (28industries) This means that I must concord data on 28industries into data on 213 4-digit Canadian SIC indus-tries The result is noisy data I thus prefer using USdata Clearly however it does not matter which I useFinally the Japanese and UK data are from the UNIDOdatabase


industry-level regressions are capturing within-plant effects rather than between-plant effects7

The US tariff concessions had no effect onemployment at the plant level but modestlyreduced employment at the industry level Thismeans that the US tariff concessions must haveforced more labor-intensive plants to contractMy student Alla Lileeva has refined this obser-vation by showing that the plant-level resultreflects the effect of pooling across exporters(for which US 0) and nonexporters (forwhich US 0) She has linked the Canadianplant-level data to data on the exporter status ofthe plant While the match precludes using mydifference-of-differences methodology she hasnevertheless been able to show that US is pos-itive for exporters and hugely negative for non-exporters Why The US tariff concessionshad the unexpected effect of encouraging Cana-dian exporters to expand their domestic opera-tions at the expense of Canadian nonexportersSince the majority of plants are nonexporterspooling across exporters and nonexporters yieldsestimates of US that are close to 0

Returning to the plant-level estimates in Ta-ble 1 row 11 excludes the plant-level controlsComparison with row 10 shows that CA or US

are unaffected by the exclusion of the plant-level controls

Rows 12ndash15 report the IV results A keyissue is the identification of variables that sat-isfy the two requirements of an instrument Themost likely candidates for valid instruments arevariables measuring the level of industry char-acteristics in 1980 For one these level char-acteristics are unlikely to be correlated withthe residuals because the latter are twice-differenced Such difference of differences arefar removed from levels For another the 1980characteristics determine the 1980 levels of pro-

tection which in turn are correlated with thetariff changes I therefore use an instrument setthat consists of 1980 log values for (1) Cana-dian hourly wages which captures protectionfor low-wage industries as in W M Cordenrsquos(1974) conservative social welfare function (2)the level of employment which captures pro-tection for large industries as in the J MichaelFinger et al (1982) high-track protection forlarge industries (3) Canadian imports from theUnited States and (4) US imports from Can-ada I also include squares and cross-products aswell as any exogenous regressors The first-stage R2s are between 030 and 040 for almostall the results in this paper

Row 12 repeats the specification of row 1 butwith the two tariff regressors instrumented CA

and US are now much larger Also US re-verses signs suggesting that the US tariffconcessions raised Canadian employment How-ever these results do not pass the Hausman test

The ldquoOverIdHausmanrdquo column reports p-values for overidentification and Hausman testsIn row 12 both the overidentification test (060)and the Hausman test (065) are above 001which indicates that the instruments are valid atthe 1-percent level and that endogeneity is re-jected at the 1-percent level Given the poorsmall-sample properties of IV estimators (CharlesR Nelson and Richard Startz 1990) I use the1-percent cut-off ie p-values below 001

Row 13 reports the IV estimates for the casewhere the US control is instrumented alongwith the two tariff concessions Comparing row13 with row 12 it is clear that endogenizing theUS control has no impact on the estimates ofCA and US Further endogeneity continues tobe rejected8

Rows 14 and 15 repeat the IV exercises ofrows 12 and 13 respectively but starting with

7 If this is not clear consider the following Let xikt besome characteristic of plant k in industry i in year t let sikt

be plant krsquos market share and let xit yenk xiktsikt be theaverage value of xikt Using obvious difference notationxit yeni xiktsikt yeni siktxikt1 ie the total industrychange can be decomposed into a within-plant change (thefirst term) and a between-plant or market-share shift change(the second term) The plant-level regressions deal withxikt and thus capture within-plant changes The industry-level regressions deal with xit and thus capture bothwithin-plant and market-share shift changes

8 As someone who has tried to build a career on theendogeneity of protection (Trefler 1993) I am surprised bythe rejection of endogeneity To investigate further I haveexperimented with a much larger set of instruments drawnfrom 1980 and 1988 characteristics of Canadian and USindustries I have also experimented with a drastically re-duced instrument set None of this makes any difference tothe conclusion that endogeneity is rejected As a result Iwill report the industry-level IV results but downplay themInterestingly endogeneity only comes into play when thedependent variable is imports See below


the plant-level baseline specification of row 10As with the industry-level results the CA andUS are much larger but endogeneity is re-jected Indeed endogeneity is easily rejected forevery plant-level specification reported in thispaper This likely reflects the fact that tariffseven if endogenous to the industry are exoge-nous to the plant

V Coefficient Magnitudes

I have not yet properly explained the magni-tudes of CA and US Since the distribution oftariff concessions is skewed it is of interest toknow the effect of the Canadian tariff conces-sions on the most impacted import-competinggroup of industries ie on the one-third ofindustries with the most negative values ofi1

CA This group has 71 (2133) industriestariff concessions ranging from 5 to 33 per-cent and an average tariff concession of 10percent The industries are listed in AppendixTable A1 For any industry i the Canadian tariffconcessions are estimated to change employ-ment by CAi1

CA log points For the most im-pacted import-competing group as a whole thischange is given by CA1

CA where 1CA is a

weighted average of the i1CA with weights

that depend on industry size (See Appendix Bfor details about the weights) It is CA1

CA

that is reported in the CA column of all thetables in this paper From row 1 of Table 1 themost impacted import-competing group as awhole experienced a 12-percent employment loss

A similar discussion of coefficient magnitudesapplies to the most impacted export-orientedgroup of industries ie the one-third of indus-tries (71 industries) with the most negative val-ues of i1

US For this group the estimated impactof the US tariff concessions on employment isgiven by US1

US where 1US is the weighted

average of the i1US US1

US is reported in theUS column of all the tables in this paper Fromrow 1 of Table 1 this group experienced a statis-tically insignificant and nonrobust 3-percentemployment loss

The ldquoTotal FTA impactrdquo (TFI) columns inthis paper present the joint effect of the tariffconcessions on manufacturing employment as awhole This effect is just

(8) TFI CA 1CA US 1

US

where 1CA and 1

US are now defined as aver-ages across all 213 industries From the TFIcolumn of row 1 in Table 1 the FTA reducedmanufacturing employment by 5 percent Thisimpact is statistically significant and quite sim-ilar across all the OLS specifications It standsin sharp contrast to Gaston and Trefler (1997)who found economically small and statisticallyinsignificant effects of the FTA The differencein conclusions reflects both the better data andthe better methodology of the current study

Employment losses of 5 percent translate into100000 lost jobs and strike me as large notleast because only a relatively small number ofindustries experienced deep tariff concessionsIndeed most of these lost jobs were concen-trated in the most impacted import-competingindustries For this group with its 12-percentjob losses one in eight jobs disappeared Thisnumber points to the very large transition costsof moving out of low-end heavily protectedindustries It reflects the most obvious of thecosts associated with trade liberalization

It is difficult to be sure whether these transi-tion costs were short-run in nature Howevertwo facts drawn from the most recent seasonallyadjusted data suggest that they probably wereshort-run costs First the FTA had no long-runeffect on the Canadian employment rate whichwas 62 percent both in April 1988 and April2002 Second Canadian manufacturing em-ployment has been more robust than in mostOECD countries For example between April1988 and April 2002 manufacturing employ-ment rose by 91 percent in Canada but fell by129 percent in the United States and by 97percent in Japan This suggests albeit not con-clusively that the transition costs were short runin the sense that within ten years the lost em-ployment was made up for by employmentgains in other parts of manufacturing

VI Labor Productivity

It would be best to examine productivity us-ing a total factor productivity (TFP) measureUnfortunately the Canadian ASM does notrecord capital stock or investment data There is


thus little alternative but to work with laborproductivity I define labor productivity as valueadded in production activities per hour workedby production workers9 I deflate using 3-digitSIC output deflators10 Table 2 reports the laborproductivity results The table has the exact

same format as the Table 1 employment resultsso that I can review it quickly As in Table1 endogeneity is always rejected11 and all theindustry-level OLS results are similar so that Ican focus on the baseline row 1 specification

From the industry-level OLS results the Ca-nadian tariff concessions raised labor produc-tivity by 15 percent in the most impactedimport-competing group of industries (t 311) This translates into an enormous com-pound annual growth rate of 19 percent Thefact that the effect is smaller and statisticallyinsignificant at the plant level (row 10) suggeststhat much of the productivity gain is coming frommarket share shifts favoring high-productivityplants Such share shifting would come about

9 Trefler (2001) extensively examined the sensitivity ofresults to alternative definitions of labor productivity Ap-pendix D of the current paper shows that the results are notsensitive to redefining labor productivity as total valueadded (in production plus nonproduction activities) perworker (production plus nonproduction workers) This def-inition does not correct for hours however it is useful inthat it is directly comparable to the way in which I amforced to define US labor productivity in yis

US (The USASM does not report value added in production activities)

10 Appendix D also shows that the results do not changewhen labor productivity is deflated by the available 2-digitSIC value-added deflators I am indebted to Alwyn Youngfor encouraging me to carefully examine the issue ofdeflators

11 The Table 2 plant-level IV results are based on aninstrument set without squares or cross-products becausethese are rejected by the overidentification tests

TABLE 2mdashDETAILED RESULTS FOR LABOR PRODUCTIVITY

Constructionof b

CanadiantariffsCA

UStariffs US

Businessconditions b

US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t



Notes The dependent variable is the log of labor productivity The estimating equation is equation (6) for the industry-levelregressions and equation (7) for the plant-level regressions The number of observations is 211 for the industry-levelregressions and 3726 for the plant-level regressions See the notes to Table 1 for additional details In rows 4 and 7 thebusiness conditions variable is omitted so that business conditions are controlled for implicitly by double-differencing yi1 yi0 In row 5 the US control is replaced by the Japan-UK control discussed in the text In row 8 the two ldquooutlierrdquoobservations with the largest Canadian tariff cuts are omitted In row 9 all nine observations associated with the automotivesector are omitted In row 11 the plant controls are omitted In rows 12 and 14 only the Canadian and US tariff variablesare instrumented In rows 13 and 15 the two tariff variables and the US control are instrumented


from the growth of high-productivity plants andthe demise andor exit of low-productivity plants

From the plant-level OLS results (row 10)the US tariff concessions raised labor produc-tivity by 14 percent or 19 percent annually inthe most impacted export-oriented group ofindustries (t 397) This labor productivitygain does not appear at the industry level(US 004 t 114) which is likely due to thefact that the US tariff concessions encouragedentry of plants that are less productive by virtueof being young (On the low productivity ofyoung plants see John R Baldwin 1995 forCanada and Andrew B Bernard and J BradfordJensen 1995 for the United States) The impor-tance of controlling for plant age can be seen bycomparing rows 10 and 11 since the latter ex-cludes the plant age control and has a lowerUS12

The last column of Table 2 looks at the totalFTA impact on all of manufacturing The plant-level numbers of row 10 indicate that the FTAraised labor productivity in manufacturing by74 percent or by an annual compound growthrate of 093 percent (t 492) The industry-level numbers are about the same These num-bers along with the 14ndash15 percent effects forthe most impacted importers and exporters areenormous The idea that an international tradepolicy could raise labor productivity so dramat-ically is to my mind remarkable

VII Import Prices and TradeCreationDiversion Implications for Welfare

Preferential trade arrangements including theFTA need not be welfare improving The liter-ature identifies two conditions which if satis-fied increase the likelihood of welfare gains fora representative domestic agent These are thattrade creation ldquodominatesrdquo trade diversion and

that import prices do not rise (Panagariya 2000Krishna 2003) This section explores theseconditions

A Trade Creation and Trade Diversion

Krishna (2003) offers a precise expression forwelfare gains in terms of the relative sizes oftrade creation and diversion Let ln misj be thelog change in Canadian imports of industry i inperiod s from region j US or j ROW (restof the world) Let isj be the correspondingchange in the Canadian tariff Krishna showsthat a sufficient condition for welfare gains is

(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0

where 08 is the share of Canadian importsoriginating from the United States13 The firstterm is proportional to a utility-relevant mea-sure of trade creation and is positive because ln mi1US i1US 0 The second term isproportional to a utility-relevant measure oftrade diversion and is likely negative because ln mi1ROWi1US is likely positive

I examine equation (9) empirically as fol-lows The first row in Table 3 reports estimatesof my standard equation (6) using Canadianimports from the United States as the dependentvariable Note that there is no US control inthis regression because it makes no sense in an

12 Another contributing factor to the difference betweenthe US at the industry and plant levels is that the US tariffconcessions encouraged Canadian plants to enter the USmarket This must reduce average productivity because newCanadian exporters are less productive than old Canadianexporters (Baldwin and Wulong Gu 2003) Expansion intothe US market therefore increases the market share oflower productivity new exporters thus reducing the industry-level productivity effect

13 To derive equation (9) start with equation (10) inKrishna iUSmiUSiUS iROWmiROWiUS where allvariables relate to 1988 Since iUS iROW in 1988 thisexpression can be rewritten as

iUS

miUS miROWiUS ln miUSiUS

1 iUS ln miROWiUS

where iUS miUS(miUS miROW) 08 is the US importshare Krishnarsquos analysis looks at a representative consumerin an economy with a single final good The generalizationto many goods is trivial as long as expenditure shares foreach good are independent of the tariff eg Cobb-Douglaspreferences In examining equation (9) empirically I ignorethe fact that Krishnarsquos miUS and miROW are compensateddemands for imports


import context The Canadian tariff concessionsraised Canadian imports from the United Statesby 54 log points I therefore set ln mi1US i1US equal to 054 The third row in Table 3reports my OLS estimates of equation (6) usingCanadian imports from the rest of the world asthe dependent variable The Canadian tariffconcessions lowered Canadian imports from therest of the world by 40 log points I therefore set ln mi1ROWi1US equal to 04014

Plugging 054 and 040 into equation (9)yields 08 13 (054) 02 13 (040) 035(t 362) Since this number is statisticallygreater than zero Krishnarsquos (2003) welfare con-dition is satisfied This conclusion is robust tothe many alternative specifications described inTables 1ndash2 Thus FTA trade creation domi-nated FTA trade diversion enough to ensure thatthe FTA improved the welfare of the ldquorepresen-tativerdquo Canadian

B Prices

A preferential trading agreement will notlikely be welfare improving if it raises prices(Panagariya 2000) Clearly the FTA is unlikelyto have raised import pricesmdashthis would re-quire either some unusual change in the strate-gic interactions between firms or a rise in tariffsagainst non-FTA trading partners More likelythe FTA reduced import prices by allowing USproducers to send larger quantities per ship-ment thus spreading fixed shipping costs over alarger number of units Fixed costs of shippingare sufficiently large that reducing them hasbeen a key focus of Canadian public policy15

Surprisingly there exists very little econometricwork on the effects of trade liberalization onimport prices J Richard Huber (1971) is a rareexception

To investigate I examine the relationship

14 Using US rather than Canadian imports Romalis(2004) finds large impacts of both the FTA and NAFTA onUS trade creation and diversion

15 See the C D Howe Border Papers series for reviewsof the public policy discussions eg Wendy Dobson(2002)

TABLE 3mdashTRADE DIVERSIONCREATION AND IMPORT PRICES

Variable

Canadian tariffs US tariffsTotal FTA

impactBusiness

conditionsAdjusted

R2OverId

Hausman ObservationsCA t US t TFI t

Canadian imports from the United StatesOLS Industry 054 467 016 216 001 083 022 024 211IV Industry 232 080 086 040 015 048 030 015 NA028 211Canadian imports from the rest of the worldOLS Industry 040 267 008 017 003 012 011 005 211IV Industry 160 054 124 048 022 047 008 004 NA075 211Canadian import pricesOLS Product 0004 020 000 4700IV Product 0073 226 000 051003 4700Canadian import quantitiesOLS Product 070 1512 005 4700IV Product 102 1268 004 087000 4700

Notes The dependent variable is indicated in bold font at the start of each block of results eg ldquoCanadian imports from theUnited Statesrdquo All dependent variables are in log changes The estimating equation is equation (6) for the industry-levelCanadian imports regressions and equation (10) for the product-level import price and quantity regressions The businessconditions variable is the same as in the Table 1 row 1 baseline specification The US control is not included because itmakes no sense in a bilateral import context CA and US are scaled as described in the notes to Table 1 An asterisk indicatesstatistical significance at the 1-percent level The ldquoOverIdHausmanrdquo column reports p-values for the overidentification andHausman tests Rejection of the instrument set or exogeneity are indicated by p-values of less than 001 Blank entries indicateOLS estimation The product-level import results use wages employment squares and cross-products as instruments Basedon the overidentification test the industry-level import results drop the squares and cross-products from the instrument setIt is thus just identified (NA)


between tariff cuts and changes in import unitvalues Both these variables are available at the10-digit Harmonized System (HS10) levelWhile unit values are difficult to interpret asprices the hope is that at this detailed level ofdisaggregation changes in unit values over theFTA period reflect changes in prices Note thatI am looking only at unit-value changes withinan HS10 item This is very different from andless problematic than the typical use made ofunit values Typically researchers draw conclu-sions from the fact that one HS10 item has ahigher unit value level than another Since unitvalues are based on actual payments net ofimport duties freight insurance and othercharges I will interpret changes in unit valuesas changes in producer prices

Canadian trade data was first collected in theHS system in 198816 Let i1j be the FTAperiod change in Canadarsquos tariff against countryj for HS10 product i Let ln pi1j be the cor-responding log import price change Since I donot have pre-FTA data on import price changesat the HS10 level ( ln pi0j) I cannot estimatemy standard equation (6) with ln pi1US ln pi0US as the dependent variable Howeverif the FTA had never been implemented oneexpects ln pi1US to have evolved in the sameway that Canadarsquos import prices from otheradvanced economies evolved I thus estimate

(10) ln pi1US ln pi1OECD

CAi1US i1OECD i

where ln pi1OECD is the simple average of the ln pi1j for the United Kingdom GermanyFrance and Japan Likewise for i1OECD

The third block of results in Table 3 labeledldquoCanadian import pricesrdquo reports the estimatesThe OLS estimate indicates that the FTA did

not raise import prices (CA 0004) Thereis modest evidence of endogeneity at the 3-percent level and the IV estimates indicate thatthe FTA reduced import prices by 7 percent forthe most impacted import-competing products

One wonders if the HS10 import pricechanges are so noisy that these results are mean-ingless Import prices are defined as import val-ues divided by import quantities so that anynoisiness in prices must come from noisiness inquantities To investigate the role of noise Ireestimated equation (10) using log importquantity changes as the dependent variable Thefourth block of results in Table 3 reports theresults The FTA raised import quantities by 70percent and the t-statistic is huge (1512) Fur-ther for the first time in this paper I obtain theexpected strong rejection of the exogeneity oftariffs Thus noise does not appear to be aproblem

To summarize two conditions increase thelikelihood that a preferential trade arrangementis welfare improving trade creation must dom-inate trade diversion and import prices must notrise Both of these condition are met in the FTAcontext

VIII Employment of Production andNonproduction Workers

I am now in a position to quickly review theresults for other outcomes The data distinguishbetween workers employed in manufacturingactivities and nonmanufacturing activities Iwill refer to these as production and nonproduc-tion workers since the distinction broadly fol-lows that used in the US ASM In particularnonproduction workers are more educatedand better paid The top block of results inTable 4 reports a limited number of specifica-tions for the employment of production work-ers My baseline industry- and plant-levelspecifications appear in rows 1 and 10 respec-tively (Row numbers match those of Table 1 sothat the reader can always remind herself of thespecification details of any row by referringback to the detailed discussion surrounding Ta-ble 1) The results indicate that the Canadiantariff concessions reduced employment by alarge amount 14 percent using industry-levelestimates (t 244) and 9 percent using

16 In matching 1988 data with 1996 data I lose 33 percentof the 1988 HS10 items There is some evidence that theloss is nonrandom in that the average tariff on the un-matched commodities is 05 percentage points lower thanon the matched commodities This reflects the fact thatmany of the unmatched commodities are in high-techindustries For example Intelrsquos introduction of the 486 CPUin 1989 quickly led to the demise of the 386 CPU (Donrsquotdate yourself by admitting you remember this)


plant-level estimates (t 258) The effects ofthe US tariff concessions are less clear Theyreduced employment by 7 percent using industry-level estimates but this is not statistically sig-nificant and virtually disappears in the plant-level estimates The total FTA impact of 8percent (industry level) and 4 percent (plantlevel) are both economically large and statisti-cally significant

Rows 4 6 and 12 present alternative speci-fications In rows 4 and 6 the business condi-tions control and the US control are excludedrespectively This does not affect the CA orUS In row 12 the industry-level IV results arereported Endogeneity is strongly rejected (p 099) I do not report the plant-level IV resultsbecause endogeneity is always strongly rejectedat the plant level

In contrast to the results for production work-ers nonproduction worker employment is esti-

mated to have been unaffected by the US tariffconcessions

Finally the ldquoSkill upgradingrdquo block of resultsin Table 4 show that there has been FTA-induced skill upgrading ie an increase in theratio of nonproduction workers to productionworkers This happened at the industry levelmuch more than at the plant level which meansthat market shares have shifted in favor ofnonproduction-worker-intensive plants Possi-bly these workers are a fixed cost that is neededto penetrate US markets

IX Earnings

Most commentators expected Canadianwages to fall in response to competition fromless unionized less educated workers in thesouthern United States Table 5 revisits thisquestion using payroll statistics Since the

TABLE 4mdashEMPLOYMENT AND SKILL UPGRADING

Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId

HausmanCA t US t TFI t

EmploymentmdashProduction workers1 Industry 014 244 007 156 008 344 037 016 0334 Industry 013 199 007 136 008 289 021 0076 Industry 016 293 008 171 009 408 037 032

12 Industry 020 128 003 017 006 160 037 016 032 05907010 Plant 009 258 003 087 004 301 017 029 004EmploymentmdashNonproduction workers1 Industry 006 071 005 079 000 002 036 007 0264 Industry 007 077 005 073 000 009 014 0006 Industry 006 079 004 071 000 012 036 026

12 Industry 001 006 011 052 005 122 036 011 025 01803610 Plant 014 302 004 119 003 172 002 015 001Skill upgrading1 Industry 011 141 010 167 008 272 047 024 0484 Industry 008 079 011 126 007 181 024 0016 Industry 012 163 010 156 008 282 047 048

12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001

Notes The dependent variable is indicated in bold font at the start of each block of results eg ldquoEmploymentmdashProductionworkersrdquo The estimating equation is equation (6) for the industry-level regressions and equation (7) for the plant-levelregressions Row numbers correspond to those in Table 1 so that the reader can refer to Table 1 for details of the specificationRows 1 and 10 are my baseline specifications See notes to Table 1 for further details including the scaling of the CA andUS An asterisk indicates statistical significance at the 1-percent level Skill upgrading is the log of the ratio of nonproductionworkers to production workers All dependent variables are in logs The number of observations in the industry-level(plant-level) regressions is 211 (3742) for production workers 212 (3539) for nonproduction workers and 211 (3489) forskill upgrading


industry-level results are robust and since endo-geneity is strongly rejected I do not report thespecifications that appeared as rows 4 6 and 12of Table 4 For all workers the tariff conces-sions raised annual earnings For example thetotal FTA impact is a rise of 3 percent at boththe industry level (t 380) and the plant level(t 564) At the plant level earnings rose forboth production and nonproduction workers Atthe industry level earnings gains were concen-trated among production workers17 I have re-

fined this observation by looking at hourlywages and hours worked by production work-ers As shown in Table 5 there are wage effectsand no hours effects These earnings and wageeffects are large in a statistical sense but smallin an economic sense For example a 3-percentrise in earnings spread over eight years will buyyou more than a cup of coffee but not at Star-bucks The important finding is not that earn-ings went up but that earnings did not go down

17 My earnings results contrast sharply with those ofGaston and Trefler (1997) and Beaulieu (2000) Gaston andTrefler found no statistically significant effect of the tariffconcessions on earnings The only effect Beaulieu finds isthe positive effect of US tariff concessions on nonproduc-

tion worker earnings (an effect I find only in the plant-leveldata not the industry-level data) Once again my improveddata and methodology means that my results supersedeolder results

TABLE 5mdashEARNINGS WAGES HOURS INEQUALITY AND OUTPUT

Variable

Canadian tariffs US tariffs Total FTA impactBusiness

conditionsUS

controlAdjusted

R2CA t US t TFI t

EarningsmdashAll workers1 Industry 005 243 003 192 003 380 034 025 020

10 Plant 004 292 004 360 003 564 017 019 003EarningsmdashProduction workers1 Industry 004 212 000 002 002 361 016 011 007

10 Plant 005 325 003 257 003 474 012 021 002EarningsmdashNonproduction workers1 Industry 001 030 001 029 000 002 018 012 008

10 Plant 004 148 006 287 003 367 011 011 001Hourly wages of production workers1 Industry 005 315 003 184 003 437 060 013 033

10 Plant 006 323 002 140 003 404 020 016 001Annual hours of production workers1 Industry 001 048 002 175 001 194 002 014 001

10 Plant 002 090 001 080 000 012 003 007 000Earnings inequality1 Industry 004 132 001 055 002 166 042 005 021

10 Plant 001 046 002 097 000 041 013 008 000Gross output per plant in production activities1 Industry 005 065 003 054 000 005 030 018

10 Plant 005 136 006 201 001 072 016 005

Notes The dependent variable is indicated in bold font at the start of each block of results eg ldquoEarningsmdashAll workersrdquoThe estimating equation is equation (6) for the industry-level regressions and equation (7) for the plant-level regressions Rownumbers correspond to those in Table 1 so that the reader can refer to Table 1 for details of the specification Rows 1 and10 are my baseline specifications See notes to Table 1 for further details including the scaling of the CA and US An asteriskindicates statistical significance at the 1-percent level Earnings inequality is the ratio of nonproduction-worker earnings toproduction-workers earnings The US control is not included in the output equations because the published data on thenumber of US plants are only available at five-year intervals All dependent variables are in logs The number of observationsin the industry-level (plant-level) regressions is 213 (3801) for the earnings of all workers 211 (3742) for the earnings ofproduction workers 212 (3526) for the earnings of nonproduction workers 211 (3738) for wages 211 (3738) for hours 211(3489) for earnings inequality and 211 (3751) for output


in response to competitive pressures from theUS South

There are a number of reasons why earningsmay have risen slightly at a time when employ-ment was falling First there may have beenend-game bargaining on the part of unions seek-ing to extract rents from nearly bankrupt firmsas in Colin Lawrence and Robert Z Lawrence(1985) To investigate I use the Canadian La-bour Force Survey which reports unionizationrates in 1996 for a classification in which man-ufacturing is divided up into 16 industries Thecorrelation of Canadian tariff concessions withunion membership rates and union coveragerates is 0016 and 0002 respectively Thusunionization does not offer an explanation ofmodestly rising earnings

Another possibility is that workers in themost impacted industries upgraded their skillspossibly through the attrition of less-skilledworkers The Labour Force Survey is the mostdetailed source of data on education by indus-try It reports education on a consistent basisback to 1988 (but not 1980) The correlation ofCanadian tariff concessions i1

CA with 1988ndash1996 log changes in average years of schoolingis 028 which supports the view that the tariffcuts were associated with educational upgrad-ing However this correlation is almost com-pletely driven by the Clothing industry Thecorrelation falls to 006 when Clothing isomitted Note of course that the Clothing indus-try is too important for an analysis of the FTAto simply be dismissed as an outlier Thuswhile there is some evidence that the earningseffect is driven in part by educational upgrad-ing this conclusion must be tentative

The explanation of modestly rising earningsbest supported by the data is seniority-basedworker attrition The Labour Force Survey re-ports current job tenure over the 1980ndash1996period Let ln Tenureis be the average annuallog change in tenure in the pre-FTA period (s 0) or FTA period (s 1) Figure 3 plots ln Tenurei1 ln Tenurei0 against i1

CA i0

CA That is it has the form of my usualdifference-of-differences estimator As is appar-ent industries that experienced the deepest tar-iff cuts (and hence the deepest employmentlosses) also experienced the largest increases incurrent job tenure The correlation is 045

The wage results point to a potential caveatfor the labor productivity results The 5-percentearnings rise associated with the Canadian tariffconcessions may in part reflect a rise in laborquality At one extreme if the earnings rise wasentirely due to increased labor quality then laborproductivity rose not by 15 percent but by 15 5 10 percent This translates into a compoundannual growth rate of 12 percent still an enor-mous number At the other extreme if produc-tivity increases drove wage increases (ie ifthere was no labor quality increase) then nocorrection to the productivity numbers isneeded

There is a presumption in the popular pressthat anything to do with globalization willworsen income inequality It is thus reassuringthat there is absolutely no evidence that the FTAworsened income inequality In the last block ofresults in Table 5 where inequality is measuredas the earnings of nonproduction workers rela-tive to production workers CA and US areeffectively 0

X What Underlies Rising Labor Productivity

To the extent that the labor productivity ben-efits of the FTA reflect gains in technical effi-ciency (as opposed to allocative efficiency) it is

FIGURE 3 CURRENT JOB TENURE CHANGES [(1996ndash1988)LESS (1986ndash1980)] VS CANADIAN TARIFF CONCESSIONS


of interest to know how this came about Thissection examines three possibilities

First plants may have moved down theiraverage cost curves To examine this I esti-mated my industry-level equation (6) for aver-age output per plant and my plant-level equation(7) for plant output The results appear at thebottom of Table 5 The industry-level CA andUS are comparable in magnitude to those esti-mated by Head and Ries (1999b) though mysignificance level is much lower18 Their find-ing of statistical significance may reflect theirdecision to work with annual changes withoutcorrecting for serial correlation The more in-teresting results are at the plant level since theseare more readily interpretable as moving alongan average cost curve The results indicate thatthe Canadian tariff concessions led the mostimpacted import-competing plants to contractby 5 percent (t 136) while the US tariffconcessions led the most impacted export-oriented plants to expand by 6 percent (t 201) These are not statistically significant re-sults Thus this is not strong evidence in sup-port of a simple scale-effects explanation oflabor productivity gains

Second the popular press reports that US-owned multinationals have been reorganizingtheir Canadian plants in order to produce fewerproduct lines each with a global mandate Thisis consistent with Baldwin et al (2002) whofind that for foreign-owned plants operating inCanada increases in exports are associated withreductions in the number of commodities pro-duced Thus plant rationalization may havecontributed to rising productivity

Third it is possible that my FTA-inducedlabor productivity gains do not extend to TFPgains However this seems unlikely since thereis little evidence of capital deepening moreintensive use of intermediate inputs or risingmarkups Specifically using my difference-of-differences methodology Trefler (2001) finds(1) no evidence of capital deepening at the3-digit SIC level (capital stock is not availableat the 4-digit level) (2) evidence of only very

modest increases in the usage of intermediateinputs at the 4-digit SIC level and (3) no evi-dence of increased markups (not a surprisegiven that the most impacted import-competingindustries are low-end manufacturing industrieswith low markups to begin with) Thus theRobert E Hall (1988) TFP calculation showsthat TFP must have risen substantially Moreexactly Trefler (2001) argues that the FTA-induced TFP changes are roughly half of thelabor productivity changes That is the TFPchanges are huge

XI Conclusions

There are many ways in which the Canada-US Free Trade Agreement provides a uniquewindow onto the effects of freer trade The FTAwas a relatively clean policy experiment un-tainted by macro shocks or financial crises Itwas an agreement between two industrializedcountries It was a reciprocal agreement whichmeans it affected exporters not just importersIn contrast most previous studies of trade lib-eralization have dealt with the unilateral tradeactions of a developing country Several strongconclusions emerged from the analysis Firstthe FTA was associated with substantial em-ployment losses 12 percent for the most im-pacted import-competing group of industriesand 5 percent for manufacturing as a wholeThese effects appear in both the industry- andplant-level analyses Second the FTA led tolarge labor productivity gains For the mostimpacted export-oriented group of industrieslabor productivity rose by 14 percent at theplant level For the most impacted import-competing group of industries labor productiv-ity rose by 15 percent with at least half of thiscoming from the exit andor contraction of low-productivity plants For manufacturing as awhole labor productivity rose by about 6 per-cent which is remarkable given that much ofmanufacturing was duty-free before implemen-tation of the FTA Third the FTA created moretrade than it diverted and possibly lowered im-port prices Thus the FTA likely raised aggre-gate welfare

The FTA is the wellspring of one of the mostheated political debates in Canada This heat is

18 Head and Ries (1999b) find CA 011 with t 308 and US 006 with t 274 (For comparability Ihave scaled their estimates)


generated by the conflict between those whobore the short-run adjustment costs (displacedworkers and stakeholders of closed plants) andthose who are garnering the long-run gains(stakeholders of efficient plants consumers andpurchasers of intermediate inputs) One cannotunderstand current debates about freer tradewithout understanding this conflict Unfortu-nately much of the academic debate has beenfragmented one set of researchers has focusedon the short-run adjustment costs of worker

displacement while another has focused on thelong-run productivity gains While this paperdoes not provide the silver bullet that makes thecase either for or against free trade I believethat it has considerably refined the question Myhope is that the results here take us one stepcloser to understanding how freer trade can beimplemented in an industrialized economy in away that recognizes both the long-run gains andthe short-run adjustment costs borne by workersand others

APPENDIX A TARIFF DETAILS

The Canadian tariff data were supplied by Statistics Canada at the 4-digit SIC level The US tariffdata were constructed as follows The 1980ndash1988 data were converted from the TSUSA classifi-cation system (approximately 10000 products) to SITC (revision 2) (approximately 800 products)using Feenstrarsquos (1996) converter It was then converted to Canadian SIC (213 industries) using aconverter supplied by Statistics Canada This converter was largely unique but where not weightsfor prorating data across SIC industries were supplied by Statistics Canada For 1989ndash1994 tariffrates the same procedure was followed but starting from HS10 rather than TSUSA For 1996 dataI converted the Census Bureaursquos ldquoUS Imports of Merchandise December 1996rdquo (CD-96-12) datafrom HS10 to SITC (revision 3) using the supplied converter I then converted the data to SITC(revision 2) using an almost 11 converter supplied by Feenstra (1996) and proceeded as with the1980ndash1988 data

Of Canadarsquos 225 4-digit SIC industries four were excluded from the analysis because ofincomplete data and another 16 were aggregated into eight categories in order to ensure consistencyof the trade and tariff data over time The aggregated industries are 1094 and 1099 1511 and 15991995 and 1999 2911 and 2919 2951 and 2959 3051 and 3059 3351 and 3359 3362 and 3369

The tariff data are defined as duties divided by imports These data are collected at the tariff-linelevel (eg HS10 after 1988) I have compared a large number of the tariff rates so derived withpublished statutory tariff rates The two tariff rate series are the same A key issue is how toaggregate the tariff-line data up to the 4-digit SIC level Since imports are the only data reported ata comparable level of disaggregation I must follow what all empirical trade researchers do andaggregate using import weights This is accomplished in the usual way as follows Consider a single4-digit SIC industry let i be an HS10 item feeding into the industry let I be the set of HS10 itemsfeeding into the industry let it be the tariff rate and let mit be the share of the industryrsquos importsaccounted for by product i My tariff rate changes have the form yeniI itmit yeniI it1mit1 For later reference yeniI (it it1)mit yeniI (mit mit1)it1

Ideally I would prefer to use fixed-weight tariffs fixed yeniI (it it1)mit1 However thiscannot be calculated because about one-third of all 1988 HS10 items disappeared by 1996(Companies often hire lawyers to have their HS10 product reallocated to a higher tariff HS10) Toget a handle on the difference between fixed and I manipulated the estimates of fixed that wereused by the Government of Canada in its pre-FTA assessment of the likely impacts of the agreement(S Magun et al 1988) To understand what I did note that most industries had their tariffs reducedto zero linearly either over five years or ten years Using Magun et al (1988) I classified 4-digit SICindustries into either the five- or ten-year category (The Magun et al study reported estimates offixed using an input-output table classification that breaks manufacturing into about 60 industries)In the formula fixed yeniI (i1996 i1988)mi1988 I set i1996 0 for five-year industries andi1996 020i1988 for ten-year industries This allows me to compute fixed

The outcome of this procedure is estimates of i1CAfixed and i1

USfixed where I am using the


notation of equation (2) Across 4-digit SIC industries the correlation of i1CAfixed with i1

CA is 098and the correlation of i1

USfixed with i1US is 097 That is my tariff rate changes are very similar to

a best estimate of fixed-weight tariff changes Not surprisingly the two tariff-change series yieldalmost identical results for estimates of CA and US Trefler (2001 Appendix 2) discusses furtheraspects of aggregation

Table A1 reports i1CA and i1

US for the most impacted import-competing industries

APPENDIX B SCALING CAAND US

AND DEFINING ldquoTOTAL FTA IMPACTrdquo

Recall that Yi1988 is the level of say employment in industry i in 1988 The industry i change inemployment over the FTA period is approximately 8(yi1)Yi1988 ie the log change times the initiallevel Multiplying by eight converts the average annual changes for the eight FTA years into a totalFTA period change The change in employment among industries in any set I is approximately8 yeniI (yi1)Yi1988 As a proportion of total employment it is 8 yeniI yi1i where i Yi1988yenjI

Yj198819 Using the fact that 8yi1ˆ 8ki1

k (k CA US) is the predicted impact of country krsquostariff concessions in industry i the predicted tariff-induced log change in employment is 8 yeniIki1

k i where I is the set of industries in the most impacted import-competing industries (k CA)or export-oriented industries (k US) Defining 1

k 8 yeniI i1k i the predicted impact reduces

to k1k which is what is reported in the tables

APPENDIX C ESTIMATION OF bis

As noted in Section IV construction of bis requires the preliminary step of estimating

1 yit i j 0

J

ij1 zt j it

I use OLS since my only criterion is to minimize in-sample prediction error This regression wasestimated separately for each industry using 1983ndash1996 data (I do not have data for 1982) Thisleaves only 13 observations for estimating seven parameters (i0 i1 and i2 are each tuples) Tomodestly increase the degrees of freedom I estimated the regression at the 3-digit SIC industry levelrather than at the 4-digit SIC industry level There is not much difference between the 3- and 4-digitbis as can be seen from the fact that on average there are only 203 4-digit industries per 3-digitindustry

Since bis is a generated regressor I reestimated all my results for the case where bi1 bi0 isan endogenous regressor in equations (6) and (7) This had no impact on the results Further tests ofmisspecification due to a generated regressor led to rejection of misspecification

Table A2 reports results for different choices of years As is apparent the results do not changesubstantially as long as the FTA baseline year is 1988 A referee has suggested that I also reportresults for the periods 1981ndash1988 and 1989ndash1996 Since the worst of the FTA adjustment happenedimmediately the use of 1989 as the FTA baseline period means that I miss at least some of theadjustment Indeed the estimated coefficients are somewhat smaller

19 There are some exceptions to this definition of i For the cases of production worker earnings and wages i is basedon total hours worked by production workers For the cases of skill upgrading and inequality i is based on total employmentFor intraindustry trade i is based on Canadian imports from the United States Otherwise if Yi1988 is a ratio then i is basedon the numerator of the ratio ie if Yi1988 ai1988bi1988 then i ai1988yenjI aj1988


APPENDIX D MEASURING LABOR PRODUCTIVITY

Table A3 reports the results for labor productivity using three alternative measures of laborproductivity The most commonly used measure of labor productivity at the industry level is valueadded per worker deflated by an output deflator This is the third measure reported in Table A3There are several defects with this measure two of which are easily addressed

The first deals with the measurement of labor input In Canada but not in the United States therehas been a strong trend towards part-time employment By not correcting for Canadian hoursmeasure 3 has a downward trend Since this trend will be spuriously correlated with the downwardtrend in tariffs the estimated effect of the FTA on productivity (CA and US) will be downwardbiased The Canadian data allow for an hours correction Unlike the US data value added isreported for production activities alone and thus can be directly compared with the data reported forhours worked Measure 1 of Table A3 reports the estimates using Canadian real value added inproduction activities per hour worked and US real value added in all activities per employee Thisis the same measure used in Table 2 As expected the estimates tend to be larger for measure 1 thanfor measure 3 (though both are large) Clearly measure 1 is preferred

The second data issue deals with deflators In Table A3 measures 1 and 3 use output deflatorswhile measure 2 uses value-added deflators Value-added deflators would have been preferable hadthe US deflator not been seriously flawed for present purposes It is at the 2-digit level (20industries) and even at this highly aggregated level there are imputations for instruments (SIC 38)and electric and electronic equipment (SIC 36) Measure 2 of Table A3 the value-added deflatedmeasure thus has serious problems This said the (CA US) based on value-added deflators are verysimilar to the (CA US) based on output deflators This can be seen by comparing measures 1 and2 in Table A3 See Trefler (2001 Appendix 4) for a detailed discussion of deflators

APPENDIX E PLANT SELECTION ISSUES

As noted in Section II subsection E my results apply to long-form plants that were in existencein 1980 1986 1988 and 1996 These tend to be large plants For example in 1988 the averagelong-form plant was 22 times larger than the all-plant average Note that the average long-formcontinuing plant was only 21 times larger than the all-continuing-plant average so that the large sizeof my plants is due to the fact that they are long-form rather than continuing per se

The available evidence suggests that long-form selection issues are of secondary importance in thecurrent context To see this I begin by noting that almost every plant in Canada receives either along-form or short-form survey so that almost the entire universe of Canadian plants is surveyedNext for the few industry outcomes available in the short-form survey (employment earningsoutput and a measure of labor productivity) the estimates of CA and US based on long-form andon long-form plus short-form plants are very similar The exception is the estimate of US foremployment It implies employment losses of 4 percent using the long-form plants and 67percent using long-form plus short-form plants Thus the conclusions from the long-form continuingplants appear to be broadly representative of all continuing plants


TABLE A1mdashTHE 71 MOST IMPACTED IMPORT-COMPETING INDUSTRIES

SIC Industry description i1CA i1

US

1131 Brewery Products Industry 0331 00123271 Shipbuilding and Repair Industry 0241 00121931 Canvas and Related Products Industry 0183 00082433 Menrsquos and Boysrsquo Pants Industry 0170 00532443 Womenrsquos Dress Industry 0162 00762491 Sweater Industry 0159 01252451 Childrenrsquos Clothing Industry 0159 00312441 Womenrsquos Coat and Jacket Industry 0157 00491993 Household Products of Textile Materials 0156 00172442 Womenrsquos Sportswear Industry 0154 00532494 Hosiery Industry 0152 00401911 Natural Fibers and Felt Processing 0150 00412434 Menrsquos and Boysrsquo Shirts and Underwear 0147 00722432 Menrsquos and Boysrsquo Suits and Jackets 0147 00652431 Menrsquos and Boysrsquo Coat Industry 0143 00792493 Glove Industry 0140 00202496 Foundation Garment Industry 0137 00291712 Footwear Industry 0127 00822612 Upholstered Household Furniture Industry 0112 00011998 Tire Cord Fabric and Other Textiles Products 0108 00472611 Wooden Household Furniture Industry 0106 00022499 Other Clothing and Apparel Industries 0103 00402581 Coffin and Casket Industry 0101 00042495 Fur Goods Industry 0097 00532444 Womenrsquos Blouse and Shirt Industry 0094 01042649 Other Office Furniture Industries 0090 00021041 Fluid Milk Industry 0089 00061991 Narrow Fabric Industry 0089 00022619 Other Household Furniture Industries 0089 00123761 Soap and Cleaning Compounds Industry 0088 00321829 Other Spun Yarn and Woven Cloth 0088 00813242 Commercial Trailer Industry 0087 00043792 Adhesives Industry 0084 00251713 Luggage Purse and Handbag Industry 0082 00732543 Wooden Door and Window Industry 0079 00391691 Plastic Bag Industry 0079 00233612 Lubricating Oil and Grease Industry 0079 00042641 Metal Office Furniture Industry 0079 00012811 Business Forms Printing Industry 0078 00161921 Carpet Mat and Rug Industry 0078 00211083 Sugar and Chocolate Confectionery 0077 00243751 Paint and Varnish Industry 0073 00362542 Wooden Kitchen Cabinets Vanities 0073 00021141 Wine Industry 0071 00303771 Toilet Preparations Industry 0070 00243993 Floor Tile Linoleum and Coated Fabrics 0070 00452721 Asphalt Roofing Industry 0069 00443791 Printing Ink Industry 0069 00172492 Occupational Clothing Industry 0066 00313542 Structural Concrete Products Industry 0066 00153021 Metal Tanks (Heavy Gauge) Industry 0066 00113029 Other Fabricated Structural Metal Products 0065 00333931 Sporting Goods Industry 0065 00101821 Wool Yarn and Woven Cloth Industry 0061 00042733 Paper Bag Industry 0061 00423243 Non-Commercial Trailer Industry 0060 00091621 Plastic Pipe and Pipe Fittings Industry 0058 00313311 Small Electrical Appliance Industry 0058 0024


TABLE A2mdashDIFFERENT CHOICES OF PRE-FTA AND FTA PERIODS

Variable

Canadian tariffs CA US tariffs US

CA t US t

Employment OLS1980ndash1986 1988ndash1996 012 235 003 0671980ndash1988 1988ndash1996 009 203 000 0041980ndash1986 1988ndash1994 013 235 000 0021981ndash1988 1989ndash1996 010 205 001 014

Productivity OLS1980ndash1986 1988ndash1996 015 311 004 1141980ndash1988 1988ndash1996 015 335 000 0041980ndash1986 1988ndash1994 017 274 001 0201981ndash1988 1989ndash1996 012 264 004 103

Notes The dependent variable is given in bold font The estimating equation is equation (6)All rows correspond to the Table 1 row 1 baseline specification except in the choice of yearsused for the difference of differences

TABLE A1mdashContinued


US

1051 Cereal Grain Flour Industry 0057 00083032 Prefabricated Portable Metal Buildings 0057 00002941 Iron Foundries 0057 00021093 Potato Chips Pretzels and Popcorn 0056 00173991 Broom Brush and Mop Industry 0055 00402792 Stationery Paper Products Industry 0054 00131052 Prepared Flour Mixes and Cereals 0054 00212819 Other Commercial Printing Industries 0052 00032799 Other Converted Paper Products 0051 00133031 Metal Door and Window Industry 0051 00322821 Platemaking Typesetting and Bindery 0051 00121012 Poultry Products Industry 0051 00173594 Non-Metallic Mineral Insulation 0049 0058

Notes This table reports 1988ndash1996 changes in tariff concessions for those industries in themost impacted import-competing group An asterisk indicates that the industry is also in themost impacted export-oriented group of industries


REFERENCES

Arellano Manuel and Honore Bo ldquoPanel DataModels Some Recent Developmentsrdquo inJames J Heckman and Edward Leamer edsHandbook of econometrics Vol 5 Amster-dam North-Holland 2001 pp 3229ndash96

Baldwin John R The dynamics of industrialcompetition A North American perspectiveCambridge MA Cambridge UniversityPress 1995

Baldwin John R Beckstead Desmond andCaves Richard ldquoChanges in the Diversifica-tion of Canadian Manufacturing Firms(1973ndash1997) A Move to SpecializationrdquoStatistics Canada Analytical Studies BranchResearch Paper Series No 179 February2002

Baldwin John R and Gu Wulong ldquoParticipationin Export Markets and Productivity Perfor-mance in Canadian Manufacturingrdquo Statis-tics Canada Analytical Studies BranchResearch Paper Series No 011 August 2003

Bartelsman Eric J and Gray Wayne ldquoTheNBER Manufacturing Productivity Data-baserdquo National Bureau of Economic Re-search (Cambridge MA) Technical WorkingPaper No 205 October 1996

Beaulieu Eugene ldquoThe Canada-US Free TradeAgreement and Labour Market Adjustmentin Canadardquo Canadian Journal of EconomicsMay 2000 33(2) pp 540ndash63

Bernard Andrew B and Jensen J BradfordldquoExporters Jobs and Wages in US Manu-facturing 1976ndash1987rdquo Brookings Papers onEconomic Activity Microeconomics 1995pp 67ndash112

Clausing Kimberly A ldquoTrade Creation andTrade Diversion in the Canada-United StatesFree Trade Agreementrdquo Canadian Journal ofEconomics August 2001 34(3) pp 677ndash96

Corden W M Trade policy and economic wel-fare Oxford Clarendon Press 1974

Currie Janet and Harrison Ann E ldquoSharing theCosts The Impact of Trade Reform on Cap-ital and Labor in Moroccordquo Journal of LaborEconomics July 1997 15(3) pp S44ndash71

Dobson Wendy ldquoShaping the Future of theNorth American Economic Space A Frame-work for Actionrdquo C D Howe Institute Com-mentary No 162 April 2002

Feenstra Robert C ldquoUS Imports 1972ndash1994Data and Concordancesrdquo National Bureau ofEconomic Research (Cambridge MA) Work-ing Paper No 5515 March 1996

Finger J Michael Hall H Keith and NelsonDouglas R ldquoThe Political Economy of Ad-ministered Protectionrdquo American EconomicReview June 1982 72(3) pp 452ndash66

Gaston Noel and Trefler Daniel ldquoProtectionTrade and Wages Evidence from US Man-ufacturingrdquo Industrial and Labor RelationsReview July 1994 47(4) pp 574ndash93

ldquoUnion Wage Sensitivity to Trade and

TABLE A3mdashSENSITIVITY TO DIFFERENT DEFINITIONS OF LABOR PRODUCTIVITY

Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol

AdjustedR2CA t US t TFI t

1 Labor productivitymdashProduction activitiesmdashHours adjustedmdashOutput deflators1 Industry 015 311 004 114 006 379 025 016 031

10 Plant 008 170 014 397 007 492 012 000 0062 Labor productivitymdashProduction activitiesmdashHours adjustedmdashValue-added deflators1 Industry 017 296 003 067 006 326 019 013 016

10 Plant 010 206 016 458 009 569 007 020 0073 Labor productivitymdashAll activitiesmdashNot hours adjustedmdashOutput deflators

1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009

Notes The dependent variable is indicated in bold font at the start of each block of results The estimating equation is equation(6) for the industry-level regressions and equation (7) for the plant-level regressions Rows 1 and 10 are my baselinespecifications as in Table 1 See the notes to Table 1 for further details including the scaling of the CA and US All estimatesare OLS An asterisk indicates statistical significance at the 1-percent level All dependent variables are in logs The numberof observations in the industry-level (plant-level) regressions is 211 (3726) for measures 1 and 2 and 213 (3801) for measure 3


Protection Theory and Evidencerdquo Journal ofInternational Economics August 199539(1ndash2) pp 1ndash25

ldquoThe Labour Market Consequences ofthe Canada-US Free Trade Agreementrdquo Ca-nadian Journal of Economics February1997 30(1) pp 18ndash41

Hall Robert E ldquoThe Relation between Price andMarginal Cost in US Industryrdquo Journal ofPolitical Economy October 1988 96(5) pp921ndash47

Harrison Ann E ldquoProductivity Imperfect Com-petition and Trade Reform Theory and Evi-dencerdquo Journal of International EconomicsFebruary 1994 36(1ndash2) pp 53ndash73

Harrison Ann E and Hanson Gordon H ldquoWhoGains from Trade Reform Some RemainingPuzzlesrdquo National Bureau of Economic Re-search (Cambridge MA) Working Paper No6915 January 1999

Harrison Ann E and Revenga Ana ldquoThe Ef-fects of Trade Policy Reform What Do WeReally Knowrdquo National Bureau of Eco-nomic Research (Cambridge MA) WorkingPaper No 5225 August 1995

Head Keith and Ries John ldquoCan Small-CountryManufacturing Survive Trade LiberalizationEvidence from the Canada-US Free TradeAgreementrdquo Perspectives on North Ameri-can Free Research Publication No 1 Indus-try Canada April 1999a

ldquoRationalization Effects of Tariff Re-ductionsrdquo Journal of International Econom-ics April 1999b 47(2) pp 295ndash320

ldquoIncreasing Returns versus NationalProduct Differentiation as an Explanation forthe Pattern of US-Canada Traderdquo AmericanEconomic Review September 2001 91(4)pp 858ndash76

Helleiner Gerald K ldquoIntroductionrdquo in GeraldK Helleiner ed Trade policy and industri-alization in turbulent times London Rout-ledge 1994 pp 1ndash36

Huber J Richard ldquoEffect on Prices of JapanrsquosEntry into World Commerce after 1858rdquoJournal of Political Economy MayndashJune1971 79(3) pp 614ndash28

Krishna Pravin ldquoAre Regional Trading Part-ners lsquoNaturalrsquordquo Journal of Political Econ-omy February 2003 111(1) pp 202ndash26

Krishna Pravin and Mitra Devashish ldquoTrade

Liberalization Market Discipline and Pro-ductivity Growth New Evidence From In-diardquo Journal of Development EconomicsAugust 1998 56(2) pp 447ndash62

Krishna Pravin Mitra Devashish and ChinoySajjid ldquoTrade Liberalization and Labor De-mand Elasticities Evidence from TurkeyrdquoJournal of International Economics Decem-ber 2001 55(2) pp 391ndash409

Krueger Anne O ldquoTrade Policy and EconomicDevelopment How We Learnrdquo AmericanEconomic Review March 1997 87(1) pp391ndash409

Lai Huiwen and Trefler Daniel ldquoThe Gains fromTrade with Monopolistic Competition Specifi-cation Estimation and Mis-SpecificationrdquoNational Bureau of Economic Research (Cam-bridge MA) Working Paper No 9169 Sep-tember 2002

Lawrence Colin and Lawrence Robert Z ldquoMan-ufacturing Wage Dispersion An End GameInterpretationrdquo Brookings Papers on Eco-nomic Activity 1985 (1) pp 47ndash106

Levinsohn James ldquoTesting the Imports-as-Market-Discipline Hypothesisrdquo Journal ofInternational Economics August 1993 35(1ndash2)pp 1ndash22

ldquoEmployment Responses to Interna-tional Liberalization in Chilerdquo Journal ofInternational Economics April 1999 47(2)pp 321ndash44

Magun S Rao S Lodh B Lavall L andPierce J ldquoOpen Borders An Assessment ofthe Canada-US Free Trade AgreementrdquoEconomic Council of Canada (Ottawa) Dis-cussion Paper No 344 1988

Nelson Charles R and Startz Richard ldquoSomeFurther Results on the Exact Small SampleProperties of the Instrumental Variables Es-timatorrdquo Econometrica July 1990 58(4) pp967ndash76

Panagariya Arvind ldquoPreferential Trade Liberal-ization The Traditional Theory and NewDevelopmentsrdquo Journal of Economic Liter-ature June 2000 38(2) pp 287ndash331

Pavcnik Nina ldquoTrade Liberalization Exit andProductivity Improvement Evidence fromChilean Plantsrdquo Review of Economic StudiesJanuary 2002 69(1) pp 245ndash76

Revenga Ana ldquoEmployment and Wage Effectsof Trade Liberalization The Case of Mexican


Manufacturingrdquo Journal of Labor Econom-ics Pt 2 July 1997 15(3) pp S20ndash43

Rodriguez Francisco and Rodrik Dani ldquoTradePolicy and Economic Growth A SkepticrsquosGuide to the Cross-National Evidencerdquo inBen S Bernanke and Kenneth Rogoff edsNBER Macroeconomics annual 2000Cambridge MA MIT Press 2001 pp261ndash325

Romalis John ldquoNAFTArsquos and CUSFTArsquos Im-pact on International Traderdquo Mimeo Univer-sity of Chicago 2004

Trefler Daniel ldquoTrade Liberalization and theTheory of Endogenous Protection AnEconometric Study of US Import PolicyrdquoJournal of Political Economy February1993 101(1) pp 138ndash60

ldquoThe Long and Short of the Canada-US Free Trade Agreementrdquo National Bu-

reau of Economic Research (CambridgeMA) Working Paper No 8293 May 2001

Tybout James R ldquoManufacturing Firms in De-veloping Countries How Well Do They Doand Whyrdquo Journal of Economic LiteratureMarch 2000 38(1) pp 11ndash44

Tybout James R de Melo Jamie and CorboVittorio ldquoThe Effects of Trade Reforms onScale and Technical Efficiencyrdquo Journal ofInternational Economics November 199131(3ndash4) pp 231ndash50

Tybout James R and Westbrook M DanielldquoTrade Liberalization and the Dimensions ofEfficiency Change in Mexican Manufactur-ing Industriesrdquo Journal of International Eco-nomics August 1995 39(1ndash2) pp 53ndash78

Wooldridge Jeffrey M Econometric analysis ofcross section and panel data CambridgeMA MIT Press 2002


evidence abounds about efficiency gains fromliberalization (eg Anne O Krueger 1997)solid econometric evidence for industrializedcountries remains scarce When I teach my stu-dents about the effects of free trade on produc-tivity I turn to high-quality studies for Chile(Tybout et al 1991 Pavcnik 2002) Turkey(Levinsohn 1993) Cote drsquoIvoire (Harrison1994) Mexico (Tybout and Westbrook 1995)and India (Krishna and Mitra 1998) amongothers I find these studies compelling but Iwonder whether they can be expected to per-suade policy makers and voters in industrializedcountries such as Canada and the United StatesWhat is needed is at least some research focus-ing on industrialized countries

The Canada-US Free Trade Agreement of-fers several advantages for assessing the short-run costs and long-run benefits of tradeliberalization in an industrialized country Firstthe FTA policy experiment is clearly defined Indeveloping countries trade liberalization is typ-ically part of a larger package of market re-forms making it difficult to isolate the role oftrade policy Further the market reforms them-selves are often initiated in response to majormacroeconomic disturbances Macroeconomicshocks market reforms and trade liberalizationare confounded Indeed Gerald K Helleiner(1994 p 28) uses this fact to argue that ldquoEm-pirical research on the relationship between to-tal factor productivity (TFP) growth and thetrade regime has been inconclusiverdquo His viewis widely shared eg Harrison and Gordon HHanson (1999) and Francisco Rodriguez andDani Rodrik (2001) In contrast the FTA wasnot implemented as part of a larger package ofreforms or as a response to a macroeconomiccrisis Second as Harrison and Revenga (1995abstract) note ldquoTrade policy is almost nevermeasured using the most obvious indicatorsmdashsuch as tariffsrdquo Tybout (2000) echoes this crit-icism My study of the FTA is particularlycareful about constructing pure policy-mandatedtariff measures

Third the FTA is not just about import-liberalizing policies It is a reciprocal agreementthat includes export-liberalizing policies aswell It should therefore be expected to induce apronounced general-equilibrium relocation ofresources out of import-competing sectors and

into export-oriented sectors I will examinethese FTA effects on a large number of Cana-dian plant and industry outcomes At the plantand industry levels the outcomes include em-ployment and earnings of both production andnonproduction workers skill upgrading earn-ings inequality hours of work plant size andlabor productivity

Fourth the FTA is a preferential trading ar-rangement Such arrangements need not bewelfare-improving I will examine the two con-ditions usually put forward as sufficientmdashatleast informallymdashfor welfare gains These arethat trade creation must dominate trade diver-sion and that import prices must not rise(Arvind Panagariya 2000 Krishna 2003) Bothconditions are satisfied

The backdrop of the FTAmdashan industrializedcountry a clean policy experiment the directpolicy lever of tariffs general-equilibrium rec-iprocity effects and the long list of outcomesincluding employment productivity and pricesmdashwill be my basis for a rigorous and detailedexamination of the short-run costs and long-runbenefits of trade liberalization

The FTA has been the subject of severalstudies since its implementation on January 11989 Gaston and Trefler (1997) found that theFTA had no effect on earnings and only a mod-est effect on employment Beaulieu (2000)found that the employment effect was primarilydriven by modest nonproduction worker em-ployment losses Kimberly A Clausing (2001)found evidence that the FTA raised US im-ports from Canada (trade creation) but did notdivert US imports away from other US trad-ing partners John Romalis (2004) found bothtrade creation and diversion The most intrigu-ing FTA study is by Head and Ries (1999b)They found that the FTA had little net effecton industry-level average output per plant(which they take as a proxy for scale) and apuzzling effect on Canadian plant exit (exitwas induced by falling Canadian tariffs andby falling US tariffs) Unfortunately noneof these papers uses plant-level data FurtherI will argue below that at least some of thesepapers (including my own) suffer specifica-tion issues that substantively mar the infer-ences drawn about the effects of the Canada-US Free Trade Agreement











itCA and it







MANUFACTURING

(In Percents)




(1) yi1 ln Yi1996 ln Yi19881996 1988

and

yi0 ln Yi1986 ln Yi19801986 1980


(2) i1k i1996

k i1988k 1996 1988



US




k (i1986









(3) yis s CAisCA

USisUS is s 0 1




(4) yis i s CAisCA

USisUS is s 0 1














US






US

yisUS bis is s 0 1

D Estimation



CA

USi1US i0

US

yi1US yi0

US

bi1 bi0 i



US yi0US






E Plant-Level Data


(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik




III The Data










1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009

















































itCA and it







MANUFACTURING

(In Percents)




(1) yi1 ln Yi1996 ln Yi19881996 1988

and

yi0 ln Yi1986 ln Yi19801986 1980


(2) i1k i1996

k i1988k 1996 1988



US




k (i1986









(3) yis s CAisCA

USisUS is s 0 1




(4) yis i s CAisCA

USisUS is s 0 1














US






US

yisUS bis is s 0 1

D Estimation



CA

USi1US i0

US

yi1US yi0

US

bi1 bi0 i



US yi0US






E Plant-Level Data


(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik




III The Data










1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009










































(1) yi1 ln Yi1996 ln Yi19881996 1988

and

yi0 ln Yi1986 ln Yi19801986 1980


(2) i1k i1996

k i1988k 1996 1988



US




k (i1986









(3) yis s CAisCA

USisUS is s 0 1




(4) yis i s CAisCA

USisUS is s 0 1














US






US

yisUS bis is s 0 1

D Estimation



CA

USi1US i0

US

yi1US yi0

US

bi1 bi0 i



US yi0US






E Plant-Level Data


(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik




III The Data










1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009

















































US






US

yisUS bis is s 0 1

D Estimation



CA

USi1US i0

US

yi1US yi0

US

bi1 bi0 i



US yi0US






E Plant-Level Data


(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik




III The Data










1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009









































E Plant-Level Data


(7) yik1 yik0

CAi1CA i0

CA

USi1US i0

US yi1US yi0

US

bi1 bi0 xik1980 ik




III The Data










1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009











































1996



1986



Constructionof b

Canadiantariffs CA

UStariffs US

Businessconditions

bUS control

yUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t











US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009














































US yi0US

with (yi1Japan yi1

UK)2 (yi0Japan yi0

UK)2








US yi0

























CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009




























































CA where 1CA is a



CA








(8) TFI CA 1CA US 1

US

where 1CA and 1















Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009
















































Constructionof b

CanadiantariffsCA

UStariffs US


US controlyUS

AdjustedR2

OverIdHausman

Total FTAimpact

CA t US t t t TFI t













(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009
















































(9) 08 ln mi1US

i1US 02

ln mi1ROW

i1US13 0





iUS


1 iUS ln miROWiUS





B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009










































B Prices







Variable


impactBusiness

conditionsAdjusted

R2OverId








CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009











































CAi1US i1OECD i















IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009













































IX Earnings



Variable


impactBusiness

conditionsUS

controlAdjusted

R2OverId





12 Industry 011 050 015 074 010 221 047 025 048 01108310 Plant 001 030 004 148 001 096 005 017 001








Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009













































Variable


conditionsUS

controlAdjusted

R2CA t US t TFI t








10 Plant 005 136 006 201 001 072 016 005








CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009













































CA i0













XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009













































XI Conclusions































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009


































































1 yit i j 0

J

ij1 zt j it
















US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009


















































US




Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009









































Variable


CA t US t






US




REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009








































REFERENCES

















Canadiantariffs

UStariffs

Total FTAimpact

Businessconditions

UScontrol





1 Industry 011 227 003 093 002 129 020 024 01910 Plant 009 219 013 407 007 554 011 013 009








































the long and short of the canada-u.s. free trade … › ~dtrefler › papers ›...

Documents